Publication number | US7092101 B2 |
Publication type | Grant |
Application number | US 10/417,066 |
Publication date | Aug 15, 2006 |
Filing date | Apr 16, 2003 |
Priority date | Apr 16, 2003 |
Fee status | Paid |
Also published as | EP1613922A2, EP1613922A4, EP1613922B1, US20040207855, WO2004094966A2, WO2004094966A3 |
Publication number | 10417066, 417066, US 7092101 B2, US 7092101B2, US-B2-7092101, US7092101 B2, US7092101B2 |
Inventors | David J. Brady, Michael E. Sullivar |
Original Assignee | Duke University |
Export Citation | BiBTeX, EndNote, RefMan |
Patent Citations (28), Non-Patent Citations (3), Referenced by (29), Classifications (20), Legal Events (4) | |
External Links: USPTO, USPTO Assignment, Espacenet | |
This invention was made with Government support under Contract No. N01-AA-23103 awarded by NIH. The Government has certain rights in the invention.
The present invention relates to methods and systems for static multimode multiplex spectroscopy. More particularly, the present invention relates to methods and systems for static multimode multiplex spectroscopy using different multi-peak filter functions to filter spectral energy emanating from different points of a diffuse source and for combining the multi-peak measurements to estimate a property of the source.
Optical spectroscopy may be used to detect and quantify characteristics of a source, such as the average spectrum of the source or the density of chemical and biological species of the source. As used herein, the term “source” refers to the object being spectrally analyzed. In determining chemical composition of a source, the spectral signature of the target species may be encoded on the optical field by mechanisms including absorption, admission, inelastic scattering, fluorescence, and wave mixing. Many sources are referred to as large etendue because they reflect or scatter light in a large solid angle over a large area. The etendue of a source is a measure of both the spatial extent of the source and the solid angle in to which it radiates. Large etendue sources are also referred to as incoherent or diffuse sources because low spatial coherence is implicit in a large solid angle radiated over a large area. An example of a large etendue source is laser-illuminated biological tissue.
One problem with measuring light radiated from large etendue sources using conventional spectrometers is that conventional spectrometers use filters that decrease optical throughput. For example, a conventional multiplex spectrometer measures light emanating from a source using elemental detectors with a narrowband color filter placed over each detector. Using narrowband color filters over each detector reduces the optical throughput of the conventional spectrometer. As a result, conventional multiplex spectrometers that utilize narrowband filters are incapable of accurately determining the optical properties of diffuse sources.
In another type of conventional spectrometer, multimodal measurements are taken in series and the measurements are combined to estimate the optical properties of a diffuse source. The spectrometers that perform such measurements are referred to as scanning spectrometers. Using scanning spectrometers to measure diffuse sources is disadvantages because such detectors require microelectromechanical and/or piezoelectric components in order to successively apply different spectral filters to the detectors, and such components are expensive and difficult to fabricate.
A further disadvantage of scanning spectrometers is that taking multiple measurements in series increases measurement time. Increasing measurement time may be undesirable for some types of measurements, such as in vivo tissue measurements.
Yet another disadvantage of scanning spectrometers is lack of intelligent spectral filters. Conventional scanning spectrometers typically capture the full spectrum of electromagnetic radiation. Capturing the full spectrum is inefficient because some measurements contain radiation bands that are not of interest to analyzing a particular source.
Accordingly, in light of the difficulties associated with conventional spectroscopy, there exists a need for improved methods and systems for multimode multiplex spectroscopy capable of accurately and efficiently measuring characteristics of large etendue sources.
The present invention includes methods and systems for static multimode multiplex spectroscopy. According to one aspect of the invention, a method for static multimode multiplex spectroscopy includes receiving spectral energy emanating from a plurality of different points of a diffuse source and simultaneously applying different multi-peak filter functions to the spectral energy emanating from the different points to produce a multichannel spectral measurement for each point. The multichannel spectral measurements for the different points are combined to estimate a property, such as the average spectrum or the chemical composition, of the diffuse source.
One capability of the invention is capturing spectral signatures from sources that radiate into highly multimodal fields. By measuring spectral projections of different modes and different points in the field and using the projections to compute weighted measures of average spectral properties of the modes, the spectral signatures of multimodal sources can be captured. An example of aggregate spectral properties that may be determined include the mean spectrum of the modes or the mean value of portions for filters on the modal spectra.
Unlike conventional scanning spectrometers, a multimode multiplex spectrometer of the present invention is static in that it simultaneously measures multimodal projections at the same instance in time. Measuring multimodal projections at the same instance in time reduces the need for moving parts and increases the speed at which spectral projections can be measured.
A multimodal multiplex spectrometer (MMS) according to the present invention analyzes a diffuse incoherent or partially coherent optical source. Such sources are “multimodal” because they are described by multiple modes in a coherent mode decomposition [1] using:
An MMS of the present invention may include spectrally diverse multiplex filters or detectors for spectral and target estimation. Spectral diversity means that different measurement values depend on a diversity of spectral channels in the field. For spectrally diverse multiplex filters, different measurement values depend on the amplitude of more than one spectral channel in the field. The filtering function in such systems contains multiple peaks. Multiplex spectroscopy generally relies on structured transformations, such as the Fourier transform of FT-IR spectroscopy. Using volume holograms, multichannel thin film filters, 3D structured photonic materials or circuits or nanostructured electronic detectors, MMS systems can be programmed to achieve arbitrary spectral responses in each detector element. Complex “matched” multichannel design is also possible for other filter technologies, for example by structured spatio-temporal sampling of two-beam interferometers.
The present invention may also include estimation of mean spectra or mean chemical densities of diffuse sources by sampling different spectral projections on distributed spatial sampling points. Conventional spectroscopy uses tightly focused beams and spatial filtering to estimate spectral characteristics of only a single mode. In principle, the spectral densities of different modes may vary. MMS spectrometers are designed such that “multiplex” measurements combine data from both spectral and modal distributions. The measurements are combined to measure average spectral or chemical densities over the modal distribution. However, it is not required to estimate the independent spectra of specific points or modes.
MMS systems according to the invention may also use randomly distributed spectral disparity from scattering of 3D structures or from nanostructure detector elements to achieve spatially distributed multiplex signals. MMS systems of the present invention may also use constrained optimization of distributed multiplex detectors to directly estimate source spectra or chemical densities.
Accordingly, it is an object of the invention to provide methods and systems for accurately and efficiently measuring spectral properties of diffuse sources.
It is another object of the invention to provide methods and systems for static multimode multiplex spectroscopy that use different multi-peak filter functions on each detector.
It is another object of the invention to provide methods and systems for combining measurements obtained by a multimode multiplex spectrometer of the invention to determine a property of a diffuse source.
Some of the objects of the invention having been stated hereinabove, and which are addressed in whole or in part by the present invention, other objects will become evident as the description proceeds when taken in connection with the accompanying drawings as best described hereinbelow.
Preferred embodiments of the invention will now be explained with reference to the accompanying drawings of which:
The present invention includes methods and systems for simultaneously obtaining multi-peak spectral measurements from a diffuse source and for combining the multi-peak projections to determine a property of the source.
According to one embodiment of the invention, spectral content may be measured and analyzed separately for each spatial point of a diffuse source. In order to obtain a spectral measurement for each point in a diffuse source, an array of elemental detectors and filters may be used. The array of elemental detectors and filters may be positioned close to the source so that spectral energy emanating from each point can be distinguished. The spectral energy emanating from each point can be separately analyzed and used to compute a property of the diffuse source, such as the average spectrum. The procedure for separately analyzing spectral energy emanating from each point of a diffuse source is referred to herein as the local model because local spectral measurements are obtained for each point.
The system illustrated in
m(r)=∫I(ν,r)h(ν,r)dν (1)
In Equation (1), I(ν,r) is the spectral intensity distribution as a function of frequency ν and radius r, h(ν,r) is a spatially localized spectral response or filter function. The measurement at a single point r of the form shown in Equation (1) is referred to herein as a spectral projection. As will be described in more detail below, each filter in filter array 102 may have a different multi-peak filter function h(ν,r). The filter functions h(ν,r) are preferably selected so that the filter functions are invertible, tailored to the expected properties of the source being analyzed, and different from each other. Exemplary filter functions suitable for use with embodiments of the invention will be described in detail below. Definitions for variables used in the equations herein are listed in the Appendix at the end of this specification.
Measurements of local projections of the power spectrum are much easier to make than measurements of the full power spectrum at each point or in each mode. Nevertheless, these reduced measurements are useful in estimating source parameters, such as the mean spatially integrated power spectrum. The mean spectrum may be defined as
where the integral is averaging is over a source area or volume A.
As an example of how Equation (1) can be used to estimate the mean spectrum {overscore (S)} (ν), it can be assumed that there exist contours in R^{2 }over which h(ν,r) is constant. Integrating along these contours, the following equation is obtained:
{overscore (S)}(ν)=∫∫h ^{−1}(ν^{−1} ,l)m(r)dl _{□} dl _{⊥} m(l)=∫{overscore (S)}(ν)h(ν,l)dν (3)
where l is a parameter along curves orthogonal to the contours. Assuming further that there exists h^{−1}(ν,l) such that ∫h^{−1}(λ′,l)h(ν,l)dl=∂(ν′−ν), {overscore (S)}(ν) can be estimated using the following equation
{overscore (S)}(ν)=∫∫h ^{−1}(ν^{−1} ,l)m(r)dl _{□} dl _{⊥} (4)
For example, for a two-beam interferometer or multichannel fiber system, one might choose h(ν,r) =cos(2πανx/c). In this case the contours of integration are lines along the y axis. And the following measurements are obtained:
Equation (4) is the easily inverted Fourier cosine transformation of {overscore (S)}(ν). In practice, this function is sampled at discrete values of x, which yields a discrete cosine transformation of the spectral density.
More generally, both the target spectrum and the measured values may be considered discretely. In this case the transformation between the target spectrum and the measurements takes the form of a linear relationship between a vector describing the target spectrum, {right arrow over (s)}, and a vector describing the measurement state, {right arrow over (m)} of the form {right arrow over (m)}=H{right arrow over (s)}. Formally, one may estimate the target spectrum as {right arrow over (s)}_{e}=H^{−1}{right arrow over (m)}, although nonlinear estimation algorithms may improve the reconstruction fidelity. In cases where the rank H is less than the number of components in {right arrow over (s)}, nonlinear techniques in combination with target constraints, such as that the components of the vector must be nonnegative or that the target source consists of a discrete number of active channels, may be needed for target estimation.
Thus, multi-peak spectral measurements combination software 106 may have access to the inverse filter functions h^{−1}(ν′,r) for each filter function h(ν′,r) implemented by array 102. Software 106 may calculate the average spectrum for each filter function by multiplying the measurement for that filter function and the inverse of the filter function and integrating over the area of the source using Equation (4). The average spectrum for each section of the source can then be added to determine the average spectrum of the source.
In considering the local model for multimodal analysis, it is not necessary to assume that I(ν,r) is confined to a plane or even a manifold. It can be assumed that present invention samples spectral projections at a sufficiently large set of such that inversion according to Equation (4) is well conditioned without requiring that the support of the sample point be compact.
Equation (1) assumes a spectrally filtered version of the field can be captured at a point. Obtaining spectrally filtered versions of a field at each point of a diffuse source may be accomplished by incorporating nanostructured electronics or non-homogeneous atomic systems in static filter array 102. For example, a quantum dot spectrometer [5] that achieves a spatially localized spectral projection will be described in detail below. In an alternate embodiment, spectral projections may be captured by focusing each spatial point of a source into a fiber and using a static interferometer on each fiber output to capture a projection. However, this approach would result in an unwieldy and complex instrument that would be difficult to manufacture. Exemplary implementations for capturing spectral projections at individual points of a diffuse source will be discussed in detail below in the section labeled “Implementations.”
In yet another alternate embodiment of the invention, rather than measuring spectral projections at each individual point of a diffuse source, it may desirable to measure projections of different modes, apply different multi-peak filter functions to each mode, and combine the measurements from each mode to estimate a spectral property of the diffuse source. Modal measurements can be taken at a location spaced from the source and do not require elemental detectors. Hence, modal-based instruments may be less complex than the elemental-based instruments describe above.
In most spectrometers, spectrally selective measures of the field are obtained via propagation through an interferometer or filter rather than by local sampling. In these systems multi-peak spectral measurements combination module 106 may analyze the projection of the field measured by detector array 104 using a modal theory based on optical coherence functions. In one embodiment, software 106 uses the coherent mode decomposition of the cross-spectral density. The cross-spectral density is defined as the Fourier transform of the mutual coherence function Γ(r_{1},r_{2}, τ) [1]
W(r _{1} ,r _{2}, ν)=∫Γ(r _{1} ,r _{2}, τ)e ^{−i2πντ} dτ (6)
W(r_{1},r_{2}, ν) is Hermitian and positive definite in transformations on functions of r_{1}, and r_{2}, by which properties one can show that it can be represented by a coherent mode expansion of the form
where λ_{n}(ν) is real and positive and where the family of functions φ_{n}(r,ν) are orthonormal such that ∫φ*_{m}(r,ν)φ_{n}(r,ν)d^{2}r=δ_{mn}.
As discussed above, the present invention may use MMS to estimate the mean spectrum of an intensity distribution on an input plane, I(ν,r). In terms of the cross spectral density, this input intensity is I(ν,r)=W(r_{1},r_{1}, ν).
An MMS is a linear optical system that transforms the coherent modes under the impulse response h(r,r′,ν). After propagation through the system the cross-spectral density is
where ψ_{n}(r,ν)=∫φ_{n}(r′,ν)h(r,r′,ν)d^{2}r′ and the functions ψ_{n}(r,ν) are not necessarily orthogonal [6]. The MMS records measurements of the form
where A_{i }is the surface area of the i^{th }detector in detector array 104 and
and
{tilde over (H)} _{i} ^{n}(ν)=∫∫φ_{n}*(r′,ν)φ_{n}(r″, ν)H _{i}(r′,r″,ν)d^{2} r′d ^{2} r″.
As with the local model, one goal of MMS is to estimate the mean spectrum, which in this case is
The spectral content of the modes is assumed to be highly correlated, and it can be assumed that λ_{n}(ν)={overscore (S)}(ν)−Δλ_{n}(ν) such that
where
assuming that
The goal of MMS design is to create a sensor such that Equation (10) is well conditioned for inversion. Thus, similar to the local model discussed above, the software 106 may estimate the average spectrum of a diffuse source using Equation (4).
An MMS system according to the present invention may also be modeled as a direct measure of chemical or biological species. Let c_{i}(r) represent the concentration of spectral species i at position r. Suppose that species i generates a spectrum s_{i}(ν). An MMS system measures spectral positions at diverse positions r. The overall spectrum at r is
Measurements take the form
where H_{i}(r)=∫s_{i}(ν,r)h(ν,r)dν. If the measurements are considered as discrete digital samples integrated over a finite spatial range (i.e.
where A_{j }is the area of the j^{th }sensor) and it is assumed that the concentration distribution as observed by the sensor represents the mean, the transformation between the concentrations and the measurements take the form
In some cases this transformation may be linearly invertible for the mean concentrations {overscore (c)}_{i}. In most cases, however, the measurements either over constrain or under constrain the concentrations. In these cases, well known algorithms as partial least squares (PLS) may be used to estimate one or more target concentrations [7–11].
The spectral projection kernels h(ν,r) should be designed to make estimation of the {overscore (c)}_{i }tractable and efficient. The exact filter design arises from the PLS algorithm. One may recursively optimize h(ν,r) in simulation using PLS to achieve maximal fidelity. Design of h(ν,r) may consist of sampling and geometric design in the case of interferometers, but is more likely to occur through hologram, thin film filter or photonic crystal design. The design methodology for these processes is discussed below.
Spectrometers may be subdivided into various classes, such as dispersive or multiplex and scanning or static [12]. A dispersive spectrometer separates color channels onto a detector or detector array for isomorphic detection. A multiplex spectrometer measures linear combinations of spectral and temporal channels. Arrays of multiplex data are inverted to estimate the spectral density. A scanning spectrometer functions by mechanical or electro-optic translation of optical properties, as in a rotating grating or a moving mirror. A static interferometer captures full spectra in a single time step by mapping wavelength or multiplex measurements onto a static sensor array. Static grating spectrometers based on linear detector arrays have been available for some time; while static multiplex spectrometers have emerged over the past decade [13–22].
Spectrometers may be characterized on the basis of many factors, including etendue and acceptance angle, throughput, spectral resolution and resolving power. The etendue is the integral of the differential product of the solid angle of emissions over the surface of the source. The etendue may be considered roughly as the input area times the acceptance angle of the spectrometer. The throughput is the photon efficiency of the instrument. The spectral resolution is the resolution of the reconstructed spectrum. The resolving power is the ratio of the center wavelength of the reconstructed spectrum to the spectral resolution. For grating spectrometers, the spectral resolution of an instrument and the etendue are proportional.
Optical fields may be described in terms of spatial and temporal modes. The modes of a system form a complete set of self-consistent solutions to the boundary conditions and wave equations within that system. Spectroscopy measures the spectral content of optical fields by measuring the mode amplitudes as a function of wavelength. In general, spectrometers employ spatial filtering to restrict the number of spatial modes in the system. This restriction is necessary because mechanisms for determining spectral content usually assume that the field is propagating along a common axis through the system. Imaging spectrometers, in contrast, independently measure the spectrum of multiple spatial modes.
Optical spectrometer design is motivated by the fact that optical detectors are not themselves particularly spectrally sensitive. Most electronic detectors have spectrally broad response over a wide range. Optical components, such as gratings, filters, interferometers, etc., preprocess the field prior to electronic detection to induce spectrally dependent features on the intensity patterns sensed by spectrally insensitive electronic devices. However, as will be described below quantum dot detectors could change this situation [5].
The following methods may be used in filter/interferometer array 102 and detector array 104 to implement the transformation of Equation (1)
In one embodiment, static filter/interferometer array 102 and detector array 104 may be implemented using a two-beam interferometer. A two-beam interferometer, such as a Michelson, Mach-Zender, Sagnac or birefringent system, separates the source with a beam splitter or polarizer and recombines it on a detector. Two-beam interferometers have long been used as scanning Fourier transform spectrometers. In this implementation, the relative optical path along the arms of the interferometer is scanned and the optical signal is measured as a function of time delay. If the transformation from a source plane to the interferometer output plane is imaging, these instruments can function as “hyperspectral” cameras in which each pixel contains high resolution spectra. Over the past decade, there has been increasing emphasis on “static” Fourier transform interferometers. In a static interferometer, the spectrum is measured in a single shot [8–13, 15, 17]. Advantages of static interferometers include reduced dependence on mechanical components, compact and stable implementation and lowered cost.
In one embodiment, static filter/interferometer array 102 and detector array 104 may be implemented using static interferometers for large etendue sources. A static two-beam interferometer captures a spectrum in a single shot by measuring the signal generated by colliding wavefronts.
Two-beam interferometers may be further subdivided into null time delay and true time delay instruments. True time delay instruments transversely modulate the beams with a lens or other imaging instruments to reproduce the field with a delay. Despite the variety of mechanisms that create two-beam interference, uniform wavefront two-beam interferometers can be described using a relatively small number of parameters in a simple model. A two-beam interferometer interferes the beam radiated from a source with a rotated, translated and delayed version of the same beam (the possibility of a change in spatial scale is discounted for simplicity.)
m(r)=Γ(r,r,0)+Γ(r′,r′,0)+Γ(r,r′,τ)+Γ(r′,r, τ) (11)
where Γ(r,r′,τ) is the mutual coherence between points r and r′ for time delay τ. The effect of the interferometer is to rotate and displace and delay one beam relative to the other. The transformation is described by
where R_{θφ} is a rotation about the center point r_{0 }and d is a displacement. In a static instrument, R_{θφ}, d′=r_{0}+d, and τ are fixed. System design consists of selecting these parameters and the sampling points and integration areas for the measurements defined by Equation (11).
As described above, an object of the present invention is measuring spectral content of spatially broadband “incoherent” sources. The coherent mode decomposition for such a source can be expressed in terms of any complete set of modes over the spatio-spectral support of the field. Using a plane wave decomposition, the mutual coherence can be modeled as
where
One goal of static MMS according to the present invention is to measure the mean spectrum, which in this case is
The quantity that a two beam interferometer measures, however, is
τ may be non-zero, but it is fixed for a given measurement.
Generally, an interferometer measures Γ(r,r′,τ) at a fixed time over a simple manifold, such as a plane. The power spectrum of the source is estimated by taking a spatial Fourier transform along one or more dimensions in the plane, which yields
where A is the extent of the sensor plane and assume that the origin of r lies on the sensor plane. k_{lmn}′=k_{lmn}·(I−R_{θφ}) and [k_{lmn}′]∥ is the component of k_{lmn}′ parallel to the sensor plane.
Any non-zero value for d′ substantially reduces the bandwidth in k_{lmn }over which Equation (16) may be used to extract the estimated spectrum of Equation (14). Since by definition a large etendue spectrometer must accept a large bandwidth in k_{lmn}, spectrometers for which the point of rotation for the interfering beams is not in the sensor plane are not well suited for implementing filter array 102 and detector array 104. Static Sagnac, Mach-Zender and birefringent spectrometers are capable of producing a center of rotation within the sensor plane and are therefore suitable for use as filter array/spectrometer 102 and detector array 104 according to the present invention. For these sensors, Equation (15) may be used to estimate {overscore (S)}(ν) to the extent that one can assume that [k_{lmn}′]∥ can be associated with |k_{lmn}|. Since spatial frequency resolution in S_{p}(u,τ) is 1/A, it can be assumed that no ambiguity results if the longitudinal bandwidth of k_{lmn }is less that 1/A. Since the spectral resolution in estimating {overscore (S)}(ν) will be c/A, for a simple two beam interferometer, spectral resolution and etendue are inversely related. A simple static interferometer, even with image transfer between the source and sensor planes and rotation in the sensor plane, cannot simultaneously maintain high spectral resolution and high etendue. The following sections consider alternative designs to overcome this challenge.
In the illustrated example, beam splitter 502 receives light rays 524 and 526 emanating from points P1 and P2 on source 522. Light ray 524 enters beam splitter 502 and is split into components 528 and 530. Similarly, light ray 526 is incident on beam splitter 502 and is split into components 532 and 534. Light ray component 528 is reflected by mirror 504 proceeds back through beam splitter 502, through optics 518 and is focused on detector array 514. Similarly, component 530 is reflected by mirror 516 and by beam splitter 502 through optics 518 and onto detector array 514. The interference of light ray components 528 and 530 produces an interference pattern for point P1 on detector array 514. Similarly, the interference of light ray components 532 and 534 produces an interference pattern for point P_{2 }on detector array 514.
According to an important aspect of the invention, the difference in distance between interference light paths for different points on source 522 preferably varies. For example, the distance traveled by light ray component 528 is preferably different from the distance traveled by light ray component 532 in reaching detector array 514. Assuming that the spectra of different points on source 522 are related, the interference patterns for the different points detected by detector array 514 can be used to estimate a property of the source 522, such as the chemical composition or the average spectrum.
The challenge of using measurements of the form represented by Equation (15) to estimate {overscore (S)}(ν)can be addressed by revised sampling strategies for Γ(r,r′,τ). The most direct approach is to sample as a function of τ. True time delays can be introduced in the field by waveguiding or by imaging. The waveguiding approach may include coupling each point in the source plane through a different fiber interferometer.
In moving from a two-beam interferometer to a fiber array, the measurements implemented by filter array/interferometer 102 and detector 104 change from continuous transform systems to discrete sampling. While a two-beam interferometer samples by integrating on pixels across a plane, more general devices consist of discrete 3D structures and sample more general space time points in Γ(r,r′,τ). A segmented two-beam interferometer is another example of a sampled system suitable for use with the present invention. Such an interferometer may include an array of static two beam interferometers, based on Sagnac or Wollaston designs. Each interferometer may use imaging optics to induce a true coarse time delay between the beams and tilt to induce a fine propagation time delay. The aperture of the interferometers is preferably matched to the etendue of the source. The main advantage of an array of two-beam interferometers is that the number of discrete devices is much reduced relative to the fiber array approach. One may view these approaches as a spectrum spanning the effective number of time delays per interferometer from one to N. The acceptance angle of the system falls as the number of time delays per interferometer increases. In view of this trade-off and the manufacturing complexity and cost of making an array of interferometers, this approach may be less preferable than other approaches. Filters may be more cost-effective to implement than sampled interferometers. Accordingly, exemplary filter implementations are described in detail below.
The filter approach to MMS seeks to directly implement measurements of the form shown in Equation (1). As in the previous section, measurements are implemented in discrete form.
where A_{i }is the area of the ith detector element. Filters 700 may include absorptive materials, as in color cameras, or interference filters. In contrast with cameras, many different filter functions h_{i}(ν) may be implemented. In order to achieve the throughput advantage of multiplex spectroscopy [23], each filter 700 integrates a broad sample of the source spectrum. Design of the filter functions for high fidelity and high resolution reconstruction of the source is weighing design problem.
The primary disadvantages of the absorptive approach are lack of spectral resolution and lack of programmability. The advantages of the absorptive approach are that the filters may in principle be very thin. The disadvantages of the interference approach are that the filters use propagation to filter and thus must be relatively thick. Interference filters may also be relatively challenging to fabricate to precise specifications and their response may depend on angle of incidence, thus limiting the etendue over which they are effective. Accordingly, four approaches to filters suitable for use with the present invention will now be described: layered media, volume holograms, 3D structured materials and absorptive media.
The simplest thin film filters consist of a layered stack of media of different refractive indices. Commonly, such filters are constructed using periodic stacks. Wavelengths and waves resonant with the filter are selectively reflected or transmitted. Limited angular acceptance is of the primary disadvantages of conventional thin film filters, but they can be optimized for high angular degeneracy. Recently, a group at MIT has shown that a thin film filter may be designed to selectively reject all light at a given wavelength, independent of angle of incidence [24–39]. This filter acts as a very high etendue wavelength selector.
For multiplex spectroscopy, a filter with high angular invariance but also broad and nearly random spectral response is preferred. Such filters have been shown to be possible through simulation.
Ideally, for a high etendue filter, the bright bands of transmission in
If the transmission across
Creation of an MMS system using multichannel thin film filters includes realizing a large number of filters with responses as structured as the response shown in
Optimized design of the layer thicknesses in a set of spectrally varying filters is expected to substantially improve the orthogonality and inversion rank of multichannel thin film filters. As shown in
Molecules emit or absorb characteristic spectra in a variety of situations. Raman spectra, which are inelastic shifts of scattered radiation due to intermolecular vibrational resonances, are particularly characteristic of molecular sources. Ethanol, for example, has Raman lines at 436, 886, 1062, 1095, 1279 and 1455 inverse centimeters relative to the excitation source.
In diffuse multicomponent environments, many spectra signals will be present. Partial least squares algorithms weight spectral components or individual measurements so to enable estimation of target densities. In designing an MMS sensor, one balances physical reliability of a filter function against the design goals. A thin film filter, for example, can be designed to pass multiple wavelengths to selectively measure components of interest based on PLS optimization. For example,
The filter illustrated in
Variation in the filter response as a function of angle of incidence is a problem with thin film systems. Spatial filtering would be needed to restrict the angles of incidence to a range consistent with the desired spectral response.
In yet another embodiment of the invention, array 102 may be implemented using a 3D filter. A 3D filter is modulates the index of refraction along all three dimensions. The simplest form of 3D spectral filter is a volume hologram, typically recorded by a photorefractive effect. Volume holograms can be extraordinarily selective spectrally, especially if they are recorded along the direction of propagation. The disadvantage of volume holograms is that they are based on very weak index modulations and that these weak modulations fall rapidly as the complexity of the hologram is increased [40]. The advantage of volume holograms is that the spatial and spectral response of the system can be precisely programmed. As an MMS system, a volume hologram recorded as a set of “shift multiplexed” [41] reflection gratings could be set to operate as an arbitrary multichannel filter on each source point.
In yet another alternate embodiment of the invention the spectral selectivity and programmability of volume holography can be exchanged for the ease of fabrication and manufacturing associated with photonic crystals or photonic glasses (quasi-random structured materials). The idea of using photonic crystals as multiplex spectral filters is particularly promising in the context of recent results on “superprism” effects [42–50]. The superprism effect yields high spectral dispersion on propagation through photonic crystals. The effective dispersion may be 1–2 orders of magnitude in excess of corresponding values for conventional dielectrics. For MMS applications, very thin samples of microcavities or gratings may be used.
Each curve in
In yet another alternate embodiment, rather than using dielectric spheres, rectilinear structures can be used to compress the modal propagation range and create spatio-spectral structure on the sensor plane.
As noted above, multichannel filter operation can be used spectrally selective absorbers, rather than inteferometric filters. Ideally, these absorbers consist of a heterogeneous set of relatively narrow band species. For example, a detector formed from an array of quantum dots may be used, as in [5]. Rather than attempting to electrically isolate individual dot channels, however, one may integrate over a selection of spectral channels. If one made detectors containing numbers of dots proportional to or less than the number of spectral channels the dots absorbed, one would expect relatively random spectral responses in each detector. An array of such detectors might be used to reconstruct the mean spectrum.
The disclosure of each of the following references is hereby incorporated herein by reference in its entirety.
Appendix | |||
r | radial coordinate in space | ||
A | area | ||
ν | frequency | ||
k | wave vector | ||
λ | wavelength | ||
c | speed of light | ||
t | time | ||
τ | time delay | ||
m(r) | measurement at a single point r | ||
I(ν, r) | spectral intensity distribution | ||
h(ν, r) | filter function | ||
{overscore (S)}(ν) | mean power spectrum | ||
{right arrow over (s)} | target spectrum vector | ||
{right arrow over (s)}_{e} | target spectrum vector estimate | ||
{right arrow over (m)} | measurement state vector | ||
H | transformation matrix | ||
W | cross spectral density function | ||
Γ | mutual coherence function | ||
ψ | mode distribution | ||
φ | orthonormal mode distribution | ||
δ | delta function | ||
c_{i} | molecular concentration of the i^{th }sample | ||
R_{θφ} | rotation operator | ||
It will be understood that various details of the invention may be changed without departing from the scope of the invention. Furthermore, the foregoing description is for the purpose of illustration only, and not for the purpose of limitation, as the invention is defined by the claims as set forth hereinafter.
Cited Patent | Filing date | Publication date | Applicant | Title |
---|---|---|---|---|
US4054389 | Sep 23, 1976 | Oct 18, 1977 | International Business Machines Corporation | Spectrophotometer with photodiode array |
US4449823 * | Feb 12, 1982 | May 22, 1984 | The United States Of America As Represented By The Secretary Of Commerce | Device for measurement of the spectral width of nearly monochromatic sources of radiant energy |
US5059027 | Apr 11, 1989 | Oct 22, 1991 | Wisconsin Alumni Research Foundation | Spatial heterodyne spectrometer and method |
US5166755 | May 23, 1990 | Nov 24, 1992 | Nahum Gat | Spectrometer apparatus |
US5245404 | Oct 18, 1990 | Sep 14, 1993 | Physical Optics Corportion | Raman sensor |
US5321265 | Jul 15, 1992 | Jun 14, 1994 | Block Myron J | Non-invasive testing |
US5421337 * | Mar 29, 1994 | Jun 6, 1995 | Massachusetts Institute Of Technology | Spectral diagnosis of diseased tissue |
US5424545 | Jan 14, 1994 | Jun 13, 1995 | Myron J. Block | Non-invasive non-spectrophotometric infrared measurement of blood analyte concentrations |
US5434412 | Oct 1, 1993 | Jul 18, 1995 | Myron J. Block | Non-spectrophotometric measurement of analyte concentrations and optical properties of objects |
US5440388 * | Aug 2, 1993 | Aug 8, 1995 | Erickson; Jon W. | Chemical analysis and imaging by discrete fourier transform spectroscopy |
US5604594 | May 18, 1995 | Feb 18, 1997 | Schablonentechnik Kufstein Aktiengesellschaft | Device and method for determining the color value of a light |
US5660181 * | Dec 12, 1994 | Aug 26, 1997 | Physical Optics Corporation | Hybrid neural network and multiple fiber probe for in-depth 3-D mapping |
US5672875 | Jun 7, 1995 | Sep 30, 1997 | Optix Lp | Methods of minimizing scattering and improving tissue sampling in non-invasive testing and imaging |
US5777736 * | Jul 19, 1996 | Jul 7, 1998 | Science Applications International Corporation | High etendue imaging fourier transform spectrometer |
US5872655 | May 5, 1997 | Feb 16, 1999 | Optical Coating Laboratory, Inc. | Monolithic linear variable filter and method of manufacture |
US5943122 * | Jul 10, 1998 | Aug 24, 1999 | Nanometrics Incorporated | Integrated optical measurement instruments |
US6002480 | Jun 2, 1998 | Dec 14, 1999 | Izatt; Joseph A. | Depth-resolved spectroscopic optical coherence tomography |
US6002953 | May 6, 1998 | Dec 14, 1999 | Optix Lp | Non-invasive IR transmission measurement of analyte in the tympanic membrane |
US6031619 * | Jul 1, 1998 | Feb 29, 2000 | Optomed Optomedical Systems Gmbh | Imaging spectrometer |
US6057925 | Aug 28, 1998 | May 2, 2000 | Optical Coating Laboratory, Inc. | Compact spectrometer device |
US6093246 * | Dec 19, 1995 | Jul 25, 2000 | Sandia Corporation | Photonic crystal devices formed by a charged-particle beam |
US6110522 * | Apr 16, 1998 | Aug 29, 2000 | Masimo Laboratories | Blood glucose monitoring system |
US6115121 | Oct 31, 1997 | Sep 5, 2000 | The Regents Of The University Of California | Single and double superimposing interferometer systems |
US6329668 * | Jul 27, 2000 | Dec 11, 2001 | Mp Technologies L.L.C. | Quantum dots for optoelecronic devices |
US6351307 | Feb 23, 2000 | Feb 26, 2002 | The Regents Of The University Of California | Combined dispersive/interference spectroscopy for producing a vector spectrum |
US6449041 | Nov 23, 1998 | Sep 10, 2002 | Lj Laboratories, Llc | Apparatus and method for measuring optical characteristics of an object |
US6476409 * | Apr 21, 2000 | Nov 5, 2002 | Canon Kabushiki Kaisha | Nano-structures, process for preparing nano-structures and devices |
US20030031434 * | Dec 22, 2000 | Feb 13, 2003 | Fajardo James C. | Method for making separable multiple core optical fibers, the resulting fiber structures, and uses thereof |
Reference | ||
---|---|---|
1 | "MicroPac(TM) Compact Spectrometer," Optical Coating Laboratory, Inc. (Santa Rosa, USA), p. 1-6 (2001). | |
2 | Douglas A. Skoog, "Principles of Instrumental Analysis, " 3rd ed., CBS College Publishing (Philadelphia, USA), p. 146-157 and 332-339, (1985). | |
3 | PCT International Search Report. |
Citing Patent | Filing date | Publication date | Applicant | Title |
---|---|---|---|---|
US7248358 | Oct 4, 2005 | Jul 24, 2007 | Plain Sight Systems, Inc. | Devices and method for spectral measurements |
US7301625 * | Jan 19, 2006 | Nov 27, 2007 | Duke University | Static two-dimensional aperture coding for multimodal multiplex spectroscopy |
US7492462 * | Jan 17, 2006 | Feb 17, 2009 | Honeywell International, Inc. | Optochemical sensor |
US7505130 * | Nov 19, 2007 | Mar 17, 2009 | Duke University | Static two-dimensional aperture coding for multimodal multiplex spectroscopy |
US7518728 * | Sep 30, 2005 | Apr 14, 2009 | Intel Corporation | Method and instrument for collecting fourier transform (FT) Raman spectra for imaging applications |
US7532320 * | Nov 30, 2004 | May 12, 2009 | Chemimage Corporation | Multimodal method for identifying hazardous agents |
US7561274 | Sep 18, 2006 | Jul 14, 2009 | Duke University | Optical spectroscopy utilizing planar spectral filters |
US7990532 | Aug 2, 2011 | Chemimage Corporation | Method and apparatus for multimodal detection | |
US8027041 * | Sep 27, 2011 | Wavefront Research, Inc. | Compact snapshot multispectral imaging system | |
US8203716 * | Aug 3, 2007 | Jun 19, 2012 | Georgia Tech Research Corporation | Tandem Fabry-Perot etalon cylindrical beam volume hologram for high resolution/large spectral range diffuse light spectroscopy |
US8351045 | Sep 26, 2011 | Jan 8, 2013 | Wavefront Research, Inc. | Compact snapshot multispectral imaging system |
US8625091 | Sep 30, 2010 | Jan 7, 2014 | Genia Photonics Inc. | Spectrometer |
US8854624 * | Dec 20, 2011 | Oct 7, 2014 | The Trustees Of Columbia University In The City Of New York | Photonic crystal spectrometer |
US8885161 | Aug 28, 2012 | Nov 11, 2014 | Spectroclick, Inc. | Energy dispersion device |
US9234799 * | Nov 5, 2008 | Jan 12, 2016 | Centre National D'etudes Spatiales | Method and device for interferometric inversion with free sampling |
US9366571 | Oct 6, 2014 | Jun 14, 2016 | The Trustees Of Columbia University In The City Of New York | Photonic crystal sensor apparatus and techniques |
US20060092414 * | Oct 4, 2005 | May 4, 2006 | Frank Geshwind | Devices and method for spectral measurements |
US20070076208 * | Sep 30, 2005 | Apr 5, 2007 | Intel Corporation | Method and instrument for collecting fourier transform (FT) Raman spectra for imaging applications |
US20070081158 * | Jan 19, 2006 | Apr 12, 2007 | David Brady | Static two-dimensional aperture coding for multimodal multiplex spectroscopy |
US20070091315 * | Sep 18, 2006 | Apr 26, 2007 | Brady David J | Optical Spectroscopy Utilizing Planar Spectral Filters |
US20070165236 * | Jan 17, 2006 | Jul 19, 2007 | Honeywell International Inc. | Optochemical sensor |
US20080106732 * | Nov 19, 2007 | May 8, 2008 | Duke University | Static two-dimensional aperture coding for multimodal multiplex spectroscopy |
US20080192246 * | Nov 30, 2004 | Aug 14, 2008 | Chemlmage Corporation | Multimodal method for identifying hazardous agents |
US20080269617 * | Oct 24, 2006 | Oct 30, 2008 | Koninklijke Philips Electronics, N.V. | Absorption and Scattering Map Reconstruction For Optical Fluorescence Tomography |
US20100110442 * | Aug 3, 2007 | May 6, 2010 | Ali Adibi | Tandem Fabry-Perot Etalon Cylindrical Beam Volume Hologram for High Resolution/Large Spectral Range Diffuse Light Spectroscopy |
US20100256938 * | Nov 5, 2008 | Oct 7, 2010 | Centre National D'etudes Spatiales | Method And Device For Interferometric Inversion With Free Sampling |
US20120206726 * | Aug 16, 2012 | The Trustees Of Columbia University In The City Of New York | Photonic crystal spectrometer | |
US20150015692 * | Jan 30, 2013 | Jan 15, 2015 | Scanadu Incorporated | Spatial resolution enhancement in hyperspectral imaging |
WO2013116253A1 * | Jan 30, 2013 | Aug 8, 2013 | Scanadu Incorporated | Spatial resolution enhancement in hyperspectral imaging |
U.S. Classification | 356/456, 356/419 |
International Classification | G01B9/02, G01J3/28, G01J3/26, G01J3/453, G01J3/36, G01J3/02 |
Cooperative Classification | G01J3/0205, G01J3/26, G01J3/36, G01J3/02, G01J3/0256, G01J3/4531 |
European Classification | G01J3/02C, G01J3/02B, G01J3/36, G01J3/26, G01J3/02, G01J3/453B |
Date | Code | Event | Description |
---|---|---|---|
May 19, 2003 | AS | Assignment | Owner name: DUKE UNIVERSITY, NORTH CAROLINA Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:BRADY, DAVID J.;SULLIVAN, MICHAEL E.;REEL/FRAME:014082/0553 Effective date: 20030425 |
Mar 10, 2010 | FPAY | Fee payment | Year of fee payment: 4 |
Mar 10, 2010 | SULP | Surcharge for late payment | |
Jan 29, 2014 | FPAY | Fee payment | Year of fee payment: 8 |